Singapore Press Holdings Head of AI Labs Ashish Verma discusses the media company’s exciting journey into newsroom AI and customer data
Even though newspapers started to go digital way back in the late 90s and early 2000s, the true digital potential of online media consumption has not yet been fully realised by many owners of traditional press.
This is starting to change. Media consumption is a highly competitive digital space. To sell its products, keep users engaged and make life easier for journalists, media organisations are looking to data, analytics and AI to drive innovations.
Earlier this year, Business of Data spoke to New Zealand-based Stuff Ltd about how it is using data to improve its content offerings for readers, and how it plans to collect data once Google disables the use of third-party Cookies in 2022.
Singapore Press Holdings is also on an exciting digital journey, having recently established its AI Labs division to develop data and AI innovations for the companies’ numerous mastheads.
Ashish Verma, who built and heads up the AI Labs team within SPH, says the objective of the team is to bring data-driven techniques into news collecting, editing and delivery, as well as provide more tailored experiences to readers and consumers of the publisher’s products.
A New Data Journey
Since its formation in December 2019, the AI Labs group has developed two major data analytics and AI-based projects, with one gaining recognition at an AI awards event in October.
Verma says the first project he set to work on was developing a customer data platform.
“At Singapore Press Holdings, we publish 17 newspaper titles in four languages. This includes our premium newspapers such as The Straits Times, Business Times, Zaobao, Berita Harian and more,” he says.
“When a reader reads one of these products, we previously wouldn’t have any real knowledge if they were also reading any of our other products, because we were not consolidating the data.”
Verma says collecting and unifying reader customer data in a platform allows the organisation to begin the process of improving customer service, retention rates, subscriptions and security.
One of the factors driving the platform’s development is Google intention to disable third-party Cookies next year.
“We will need to rely on our own data platform now,” Verma says. “This is actually a better opportunity and situation for us because these are our products.
“We make the content the customers read, and we can use the data we collect, with the customers’ consent, for the specific purpose of improving the products and the customer experience, without that data being shared with any other company.”
Another important use case the customer data platform serves concerns paid subscriptions for premium access, both in terms of sign up and retention.
“For our readers who are not yet subscribers, we need to determine which point of time we should introduce the paywall to them,” Verma says.
“By using our AI model and propensity tool, we are getting insights from the customer that helps us decide when to show the paywall and increases the chance of conversion.
“We also use this approach to gain insights from our subscribed customers. The tool can make us aware if they are suddenly reading less so we can then deliver them some value and retain them as premium customers.”
AI in the Newsroom
The second major project Verma and his team have embarked on has this year seen them introduce AI and analytical tools into the newsroom to make journalists’ lives easier.
“We are doing this in three ways, with tools for news gathering, AI in news production and AI in news distribution,” Verma says.
In news gathering, SPH is running a program that sources articles that are heavily trending around the web, drawing from social media and news aggregators like Google News.
“These stories are consolidated and sorted on a portal provided to journalists. It helps to inform them on which news topics might be going viral. This at least makes journalists aware of what potential stories are really gaining traction,” Verma says.
In news production, Verma says his team have developed and are now testing an AI-driven virtual newscaster that can complement the news websites and print editions with rolling 24/7 coverage.
“The synthetic newsreader can be trained to deliver the news in different languages and also language variations,” Verma says. “In Singapore, the version of English spoken can often be Singlish, it’s slightly different to Western English. We’ve trained the AI newsreader to speak this way. It’s an exciting product we are working on rolling out soon.”
The third newsroom AI component Verma and his team have been working on is a tool that ingests news articles and can summarise the key points and angle of a piece. Titled SPH Robbie, it is designed to help reporters edit their articles for succinctness and clarity, as well as provide video presenters with ready-to-go story summaries.
The project has gained recognition after winning an award in the Best Use of AI/Automation for Innovative Capability category at CIO Academy Asia’s Inspire Tech Awards 2021 in October.
“It’s an AI-powered product with a nice, simple UI that, based on the requirements of the journalist, can perform multiple summarisations modelling, with sentiment analysis built in as well,” Verma says.
“We can choose the extent we want a piece summarised and the kind of summarisation. There are four different types of summary outputs, one summary model gives a different style than another, so the journalist picks from a drop-down which is most suitable to them.”
Just the Beginning
Verma has spent more than 20 years in the software industry, having worked for Barclays, Walmart and Adobe as well as becoming increasingly involved in developing AI solutions more recently.
Having joined SPH with the goal of driving new data and AI innovations into the news media and already seeing success, Verma says the news group’s data transformation journey is just at the beginning.
“When I joined SPH there was no AI Labs, I started that and we’re really helping the digital transformation journey,” he says.
“Starting with data consolidation and the building of the customer data platform. That single point of information provides a common source of truth and we delivered the first version of the product in just 9 months.
“There is great potential for use cases to come from the consolidation and management and analysis of data,” he says. “We are still early in the journey but it’s a very exciting space to be working in.”